AI and Venture Capital: How to Invest?
Aileen Lee is the excellent leader of Cowboy Ventures. In a week when OpenAI raised $6 billion in new equity financing plus another $5 billion in loan facilities, she asked whether startups have a chance in the AI game.
Another way of posing her question is whether AI is a game that small venture funds, unable to write minimum checks of $250m (according to reports), can play.
Foundation model AI is not a game for typical venture capital funds.
Aileen’s answer is optimistic. She points out that previous platform shifts (as AI seems to be) produced opportunities higher up the stack in both the business and consumer realms. The Internet and Mobile are used as examples. She does caveat that with:
That said - the enhanced ‘intelligence’ of AI offers a different ability to replace knowledge workers versus prior platform shifts. This might drive a different and earlier B2B adoption curve for AI applications vs B2C adoption. And, the current compute, cost and data considerations for AI might make iterating on ideas more challenging than developing and A/B testing ideas on the web or mobile.
But what if the optimism for a scenario like the web or mobile is wrong? What if AI foundation model companies can collapse the stack? That is to say, what if business logic and consumer experiences are both subsumed inside the new platform? The idea is not as far-fetched as it sounds, although we are not close to it today. Such an outcome would require agents to be able to take action using APIs and perhaps Apps. That seems close to happening.
If you are like me, you already use ChatGPT for multiple tasks. Almost all of them would have required numerous steps before now. And none of them are foundation model tasks. They are tasks in my work or personal life that ChatGPT can handle. I would guess hundreds of thousands of such functions are already being used at work, in education, in research, and at home.
The strict separation between platform and service or platform and application that was normal in prior technical eras might no longer apply. If that is accurate, the vertical applications the Cowboy essay is looking for might not arrive. Venture investors must look elsewhere for the next opportunity to invest in universal human experiences.
Confession: I hope Aileen is right.
For that to happen, builders need to act now. My experience of seeing my 17-year-old son use AI as a teaching aide persuades me that schools, colleges, and other educational environments will change fast.
Teachers will survive if they can morph into AI-aware mentors who guide learning. But not if they insist on being the only source of knowledge or the only reliable grade-scoring source.
Knowledge dissemination and scoring are too easy for AI to take over. But will OpenAI transform education? Or will it be a startup? Aileen asks for patience to wait for the moment. But my guess is the opposite. Opportunities must be seized now, or the foundation models will, like a virus, spread into every open opportunity, leaving no space for startups. It feels like the fastest technical penetration ever is happening under our noses.
领英推荐
My very smart colleague Rob Hodgkinson articulates that venture dry powder is being patient until better opportunities arise in his piece titled?‘Early can be tantamount to being wrong.’ He and Aileen are somewhat aligned.
In truth, the unsatisfactory answer for VC is to encourage patience. If there is indeed an opportunity for startups to displace the incumbents in the AI landscape, the power law vintage(s) for AI are ahead for VCs. Maybe this comes from agentic software? Or synthetic data? Or investments in vertical SaaS companies with proprietary data yet to be hoovered up into AI systems? Or something else entirely?
It is also true that too late can close all doors. And if the new platform tends to spread its impact like a virus or like a flood via APIs, then time is not on the side of those who wait. I can imagine the foundation models doing everything on the list.
Rob’s last sentence:
And wait. For VC’s time in the AI sun is coming. It just hasn’t arrived yet.
That is probably right, but very temporarily. At SignalRank, we realize that all ideas are provisional and discussion is permanent. So, this is not an unusual disagreement.
I believe that the dramatic uses of AI to transform processes and human behaviors are accelerating, and the entrepreneurs focused on the opportunities are not arriving as fast as the platforms are maturing. Acting fast is essential, so patience is not the key. Speed and focus are more important. Why do I believe this?
Because AI has properties unlike the Internet or mobile phones, it is a theoretical replacement for every interface and layer. The Internet has websites and web services to access data and information. So did Mobile, using apps. But AI requires no new interfaces. Indeed, it seeps into them (like Apple’s iPhone). And it can replace layers of logic. AI is more than a platform. It is a ubiquitous intelligence layer capable of addressing anything trained on and in any user interface.
But Rob is right about dry powder. Venture funds are allocating capital to companies at the lowest pace for a long time. The Q3 reports from Carta and Crunchbase point to the stagnant ecosystem outside of large AI deals.
New platforms are usually a match made in heaven for venture funds. This time, it may be a match made in hell. But like I said, I hope I am wrong, and entrepreneurs acting fast to address opportunities can make me wrong. Education seems a low-hanging fruit.
Hat Tip to this week’s creators: @aileenlee, @readmaxread, @emollick, @thomas_coatue, @ajkeen, @rpi, @mattturck, @inafried, @Samirkaji, @alex, @geneteare, @CartaInc, @rhodgkinson, @eringriffith
Contents